Table of contents
Back to resources
The New AI Superskill: Management | In The Loop Episode 46
.png)
Published by
Published on
February 26, 2026
February 26, 2026
Read time
5
min read
Category
Knowledge work is being radically transformed by tools like Claude Code, now being reimagined for knowledge workers with Cowork. We're entering an era where managing AI tools that do work for you matters more than using the tools themselves.
This shift brings brand new skill sets, ones that will be critical for you at work and for your colleagues. January's hiring and firing data in the U.S. looked painful: the worst since 2009, with companies openly saying they're removing people as a result of AI.
In this episode, I break down the exact skill sets I believe are critical to learn (judgment, taste, delegation), along with frameworks for deciding when to hand work off to AI, and what you can practically do about this change.
Management is the new superpower
Ethan Mollick did a fantastic piece on all of this. For those who don't know him, Mollick is an AI researcher and professor at the Wharton School of the University of Pennsylvania who puts out brilliant work online. I'd highly recommend checking him out.
His answer to "what skills matter now?" was simple: management. Not management in the sense of sitting in meetings and reviewing slide decks. Management as in describing exactly what you need, giving effective and clear feedback, and designing new ways of evaluating the work you get back from agents. If you're good at those things, you're going to work with agents very well.
What I loved about his piece is the nuance he adds. Management has always assumed scarcity. You delegate because you can't do everything yourself, and because talent is limited and expensive. AI completely flips this. Talent is abundant and cheap. What's scarce is knowing what to ask for and how to ask it.
Senior product managers at Google mapped out skill sets they believe are now critical to a modern worker:
- Shaping a problem: turning vague objectives into precise, testable subtasks
- Curating context: deciding what information matters for a given subtask and only providing that
- Creating documentation and systems so you can get 90% of the way there on a single attempt
- Recognizing quality
- Orchestrating many agents, managing sequential versus parallel tasks
- Understanding guardrails and ensuring agents perform well
- Knowing when to intervene with an agent or when to start over
- Exercising judgment about your toolset: when to use an agent, when to use a simple model, and when to just do it yourself
These skills aren't new. They apply to AI and there's a learning curve, but at their core, they're management fundamentals.
The CEO of Shopify, Tobi Lütke, is a fantastic case study. His GitHub commits have gone from 94 in 2024 to 833 in 2025 and over 950 in the first 45 days of 2026 alone. He's clearly not writing all that code himself. He's the CEO of a massive company. But he's orchestrating and managing agents to write code and get it into the codebase. AI is turning a CEO back into a builder because talent is now cheaper and more readily available. What matters is your ability to instruct it.
We saw something similar from Microsoft CEO Satya Nadella. At LinkedIn, they used to have separate designers, front-end engineers, back-end engineers, senior architects, and product managers. They've rolled all those skill sets into a single role they call a "full-stack builder". Someone who can have an idea, shape it, understand the problem they're solving, build much of it, and get it to a place where others can review, without relying on 18 meetings and documents.
Taste, judgment, and curiosity
Another emerging skill set is taste. Taste is about selection. Recognizing quality among many different options or directions. If AI generates ten versions of something, which one do you go for? Which one is actually good?
Judgment, on the other hand, is about weighing tradeoffs and evaluation, especially when you don't have complete information. You don't necessarily know the end direction or outcome you're heading toward. Decision-making integrates taste, judgment, and evaluation together. Then management becomes the coordination of all these decisions across many agents and tools to get more work done than ever before.
Curiosity is arguably the most desirable skill right now. You could have extraordinary taste, but without agency and curiosity, you won't create anything. You need to be constantly asking: is there another direction to take? Can I prompt it to explore this path? If it doesn't get me anywhere, that's fine. But I was curious enough to try. And then the judgment and taste to know when it's good enough to ship.
Someone could have extraordinary taste but no agency (the ability to make a decision) and therefore can't create anything. You could have plenty of agency but awful judgment, moving fast in the wrong direction. You could have all the curiosity in the world but without the ability to follow through and take action, nothing happens. All these skills combined become extremely important to develop.
We've seen these sentiments from leaders across the tech world. Paul Graham, co-founder of Y Combinator, wrote: "When anyone can make anything, the big differentiator is what you choose to make." Greg Brockman, president of OpenAI, declared that taste is now a core skill. And Cloudflare CTO Dane Knecht stated that in 2026, taste is the engineering differentiator.
You can be a full-stack builder in whatever career path you're on. Someone who can produce more than anyone else. But it requires developing these skill sets to be trusted in your organization to go and do those things.
A framework for delegating to AI
Ethan Mollick believes delegation comes down to three variables:
- Human baseline time: How long would this task take you to complete?
- Probability of success: How likely is AI to give you a good output?
- AI process time: How long would it take to write the prompt, get the answer, and evaluate the output?
On the surface, it seems straightforward. But getting those three variables right requires deep understanding of the tools, how to manage agents, what they're good and bad at, and having stress-tested those different avenues yourself.
The reason I'm emphasizing these new skill sets and frameworks is the wave of layoffs now hitting the U.S. Employers announced over 108,000 layoffs in January, a 118% increase year over year, according to Challenger, Gray & Christmas. Hiring intentions fell to their lowest level since records began in 2009. AI was cited in over 7,600 job cuts, and although that represents only 7% of the total, I believe it understates the impact AI is starting to have on the job market.
Closing thoughts
We're witnessing a shift from execution to direction. This is playing out most visibly in programming and development, but it's coming for everyone. The skills that matter now are taste, judgment, agency, curiosity, and the ability to coordinate work across many agents.
These are the sorts of capabilities people often didn't develop in themselves because they were too busy executing. Now, everybody is becoming a manager. The economic reality of this change is uneven, and in places it's going to be unfair. Layoffs in this next period are likely structural, not temporary.
But there are skills everyone must learn and demonstrate to navigate these difficult times. Nobody can fully predict what work looks like when everybody's a manager. Those who will thrive are those who can describe exactly what they need and make the perfect judgment call on when it's good enough to ship.
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.png)
.jpeg)
Book a demo today.
